Video Forgery Detection: A Comprehensive Study of Inter and Intra Frame Forgery With Comparison of State-Of-Art

Abstract

Availability of sophisticated and low-cost smart phones, digital cameras, camcorders, surveillance CCTV cameras are extensively used to create videos in our daily life. The prevalence of video sharing techniques presently available in the market are: YouTube, Facebook, Instagram, snapchat and many more are in utilization to share the information related to videos. Besides this, there are many software which can edit the content of video: Window Movie Maker, Video Editor, Adobe Photoshop etc., with this available software anyone can edit the video content which is called as “Forgery” if edited content is harmful. Usually, videos play a vital role in terms of proof in crime scene. The Victim is judged by the proof submitted by the lawyer to the court. Many such cases have evidenced that the video being submitted as proof is been forged. Checking the authentication of the video is most important before submitting as proof. There has been a rapid development in deep learning techniques which have created deepfake videos where faces are replaced with other faces which strongly made a belief of saying “Seeing is no longer believing”. The available software which can morph the faces are FakeApp, FaceSwap etc., the increased technology really made the Authentication of proofs very doubtful and un-trusty which are not accepted as proof without proper validation of the video. The survey gives the methods that are capable of accurately computing the videos and analyses to detect different kinds of forgeries. It has revealed that most of the existing methods are relying on number of tampered frames. The proposed techniques are with compression, double compression codec videos where research is being carried out from 2016 to present. This paper gives the comprehensive study of techniques, algorithms and applications designed and developed to detect forgery in videos.